transcription - CiteSeerX

The EMBO Journal vol.1 1 no.9 pp.3337 - 3346, 1992

Phosphorylation of CREB affects its binding to high and low affinity sites: implications for cAMP induced gene transcription Mark Nichols, Falk Weih, Wolfgang Schmid, Carol DeVack, Elisabeth Kowenz-Leutz, Bruno Luckow, Michael Boshart and Gunther Schutz Institute of Cell and Tumor Biology, German Cancer Research Center, Im Neuenheimer Feld 280, D-6900 Heidelberg, Germany Communicated by G.Schutz

Cyclic AMP treatment of hepatoma cells leads to increased protein binding at the cyclic AMP response element (CRE) of the tyrosine aminotransferase (TAT) gene in vivo, as revealed by genomic footprinting, whereas no increase is observed at the CRE of the phosphoenolpyruvate carboxykinase (PEPCK) gene. Several criteria establish that the 43 kDa CREB protein is interacting with both of these sites. Two classes of CRE with different affinity for CREB are described. One class, including the TATCRE, is characterized by asymmetric and weak binding sites (CGTCA), whereas the second class containing symmetrical TGACGTCA sites shows a much higher binding affinity for CREB. Both classes show an increase in binding after phosphorylation of CREB by protein kinase A (PKA). An in vivo phosphorylation-dependent change in binding of CREB increases the occupancy of weak binding sites used for transactivation, such as the TATCRE, while high affinty sites may have constitutive binding of transcriptionally active and inactive CREB dimers, as demonstrated by in vivo footprinting at the PEPCK CRE. Thus, lower basal level and higher relative stimulation of transcription by cyclic AMP through low affmiity CREs should result, allowing finely tuned control of gene activation. Key words: cAMP response elements/CREB/DNA binding/ in vivo footprinting/PKA protein phosphorylation

Introduction Cells respond to changes in their environment by transmitting external signals through several internal pathways, eliciting immediate alterations by changes in enzyme activities or later responses by reprogramming gene expression. One well characterized signal transduction pathway acts through protein kinase A (PKA). An increase in the level of the second messenger cyclic AMP (cAMP) activates PKA, resulting in phosphorylation of numerous protein substrates (reviewed in Edelman et al., 1987), leading in some cases to changes in expression of target genes (reviewed in Roesler et al., 1988; Montminy et al., 1990). The study of genes whose transcription is cAMP responsive revealed a common cAMP response element (CRE) TGACGTCA (Deutsch et al., 1988a,b). A 43 kDa protein termed CREB binds to this sequence (Montminy and Bilezikjian, 1987) and leads to increased activation of transcription after phosphorylation by PKA (Yamamoto et al., 1988; Gonzalez and Montminy, Oxford University Press

1989). A number of proteins have been identified which are related to CREB in their DNA binding domain and which recognize identical or closely related DNA sequences (Hardy and Shenk, 1988; Hai et al., 1988, 1989; Gaire et al., 1990; Hurst et al., 1990 and references therein). All appear to be leucine zipper proteins which bind DNA as dimers (Landschulz et al., 1988) and CREB has been shown to be present as a dimer in solution (Dwarki et al., 1990). Many of the leucine zipper proteins have been shown to heterodimerize with various other members of their class (Benbrook and Jones, 1990; Hai and Curran, 1991), however, CREB is relatively selective in that it has been reported to heterodimerize only with ATF-1 (Hurst et al., 1991) and CREM (Foulkes et al., 1991). CREB and CREMT are the only members of the CREB/ATF family which have been shown directly to mediate cAMP induction of transcription (Gonzalez and Montminy, 1989; Hurst et al., 1991; Foulkes et al., 1992). The tyrosine aminotransferase (TAT) gene contains a functional CRE 3.6 kb upstream of the transcription start site as an essential component of a liver-specific enhancer (Boshart et al., 1990; Weih et al., 1990). Protein binding to this site depends on the state of phosphorylation as revealed by genomic footprinting: increased intracellular cAMP concentration led to an increase in protein binding (Weih et al., 1990), and inhibition of PKA by overexpression of the regulatory subunit RIca results in a virtual disappearence of this DNA binding activity (Boshart et al.,

1991). We have focused on purification and characterization of the cAMP-induced DNA binding protein from rat liver which interacts with the CRE of the TAT gene (TATCRE) in vivo. Since PKA was reported not to stimulate binding of CREB to the somatostatin CRE (Montminy and Bilezikjian, 1987; Yamamoto et al., 1988), we anticipated that a different member of the CREB/ATF family would recognize the TATCRE in a phosphorylation-dependent manner. However, we show that phosphorylation of CREB by the catalytic subunit of PKA increases binding to several CREs, including the TATCRE. In hepatoma cells, genomic footprinting revealed that binding at the TATCRE, a weak CREB binding site, increased after cAMP stimulation, whereas for the PEPCK gene which is also cAMP responsive, the high-affinity CREB binding site showed no further increased binding after cAMP stimulation. Hence, the basal and stimulated PKA activity levels within the cell are very important in determining the degree to which CRE enhancers are bound by CREB and activate transcription.

Results Two separable binding activities interact with the

TATCRE Genomic footprinting studies showed that binding activity at the TATCRE in vivo correlates with PKA activity (Weih 3337

M.Nichols et al.

et al., 1990; Boshart et al., 1991). To characterize this binding activity, rat liver extracts were fractionated into three specific binding activities, A, B and C (Figure lA), as characterized earlier in bandshift assays using crude nuclear extracts of hepatoma cells (Weih et al., 1990). The B and C complexes are very closely related by all criteria of analysis and are referred to as BC. Methylation interference and bandshift analyses have shown that both A and BC proteins bind to the TATCRE, but not to the mutant version (MUTCRE), which has three substitutions at nucleotides contacted by protein in vivo (Weih et al., 1990). The three MUTCRE sequence changes were also shown to be sufficient

B ^r

^

"

Z

3!sw

...

Fig. 1. Two separable factors interact with the TATCRE. Rat liver nuclear extracts were applied to DEAE-sepharose and proteins were eluted with a 50-500 mM KCI gradient. Aliquots of fractions were assayed for binding activities in bandshift assays with the TATCRE oligodeoxynucleotide (A). The BC activity elutes at - 100 mM KCI from DEAE-sepharose (lanes 4-6). The A activity elutes at -200 mM KCI (lanes 7-11). Lane 1 shows the starting material activities. A non-specific binding protein complex is marked by NS and binds equally well to the TATCRE, SOMCRE and MUTCRE DNAs (not shown). In (B) fractions from the DEAE-sepharose, containing primarily the BC or the A activity, were treated with the catalytic subunit of protein kinase A (+) prior to the bandshift assays.

::

to inactivate the enhancer in vivo (Boshart et al., 1990; Weih et al., 1990). Since the BC complex was more abundant in cells following cAMP induction, in contrast to the A complex (Weih et al., 1990), we anticipated that this binding activity may correspond to the cAMP-inducible binding activity

observed in vivo. The BC activity recognizes both the CRE of the TAT and somatostatin genes while the protein forming the A complex recognizes only the TATCRE (data not

shown). BC binding to the TATCRE can be increased by treatment with the catalytic subunit of protein kinase A We tested directly the effects of the catalytic subunit of PKA on the bandshift activities, after fractionating rat liver nuclear extracts using DEAE-sepharose or S-sepharose. We found that the shifted band due to binding of BC to the TATCRE was altered in mobility and intensity after PKA treatment, whereas the A binding activity was not affected by PKA (Figure 1B). This shows a direct link between PKA and the BC binding activity, which is not apparent for the A protein(s). Hence, BC is most likely to be the binding activity responsible for the cAMP-inducible footprint in vivo. Each of the two separable activities (A and BC, Figure 1) was purified independently using native DNA cellulose and specific TATCRE oligodeoxynucleotide chromatography. CREB and the BC proteins are both recognized by antibodies to a short peptide of CREB In order to define the relationship between the BC activity from liver and the CREB factor (Yamamoto et al., 1988), we also purified CREB from rat brain. To determine the antigenic relatedness of CREB and the BC protein from liver, polyclonal antibodies against CREB were generated using a peptide representing amino acids 137-150 of rat brain CREB as antigen (Gonzalez et al., 1989). In Western blot analyses, the anti-CREB antiserum reacted with a single

.

Fig. 2. Anti-CREB antibodies react with a 43 kDa protein in the BC fraction from liver which comigrates with CREB from brain. In (A) proteins from liver and brain of rats were included in Western blot analysis using anti-CREB antibodies generated to a short peptide sequence from CREB. Samples include 150 jig protein from FTO2B extract, 50 jtg of a partially purified heparin-sepharose fraction (HEP) and samples of oligodeoxynucleotide affinity purified A, BC and brain CREB fractions (- 1, 1 and 0.2 stg protein, respectively). Positions of marker proteins from Sigma are shown at the left. CREB is 43 kDa relative to egg albumin (45 kDa). In (B) affinity purified BC of liver was preincubated with preimmune serum (PI) or antiserum (I) to CREB, prior to the addition of labelled DNA in a standard bandshift reaction. The CREB-antibody-DNA complex is marked by (*). Using concentrated antibodies (ab) (C), separated B and C forms of the BC fraction were completely immunoshifted when bound to the TATCRE (lanes 1-4). The BC fraction bound to the consensus SOMCRE was immunoshifted (lanes 5-6), as was the purified brain CREB bound to the SOMCRE (D).

3338

CREB binding to high and low affinity CREs

protein migrating at 43 kDa in both the BC and CREB fractions (Figure 2A). This protein was also detected as the most abundant specific immunoreactive protein in nuclear extract from FT02B cells (a hepatoma cell line that expresses TAT) and a partially purified BC fraction from liver. The antibodies did not recognize protein from the purified A fraction. Therefore, BC and CREB share specific antigenic determinants corresponding to amino acids 137-150 of CREB. In addition, bandshifts performed with purified BC from liver in the presence of antiserum show a specific immunocomplex, denoted by *, most likely corresponding to a BC -antibody -DNA complex, which is not present in bandshift assays containing the pre-immune serum (Figure 2B). Hence, the antibodies recognize native BC protein bound to DNA. To test the relationship between complexes B and C, two fractions that were almost devoid of the other form were tested in immuno-bandshift assays using concentrated anti-CREB antibodies (see Materials and methods). A complex of slower mobility with proteins comprising either the B or C complex from liver (showing relatedness of the two forms; Figure 2C), as well as the purified CREB from rat brain (Figure 2D) was obtained, indicating that the proteins each share the CREB epitope. The shift in mobility by the affinity-purified antibodies (Figure 2C and D) appears to be quantitatively complete as DNA binding is not inhibited. These results were observed with the BC fraction bound to the TATCRE or to the consensus CRE of the somatostatin gene (SOMCRE). As also shown in Western blot analysis (Figure 2A), the antiCREB antibodies do not recognize the A protein(s) in immuno-bandshift assays (data not shown). Thus, BC from liver and CREB may be related or identical, while the A protein does not appear to be related to CREB. Phosphorylation of CREB purified from rat brain increases binding to the TATCRE To test whether the increased binding seen in partially purified fractions from liver after PKA treatment is brought about by CREB homodimers, we tested highly purified CREB protein from brain, consisting predominantly of a 43 M

2 3

A. 116 kD 98 kD 66 kD

pror

B.

~~~~~~~~~~~~~~~~~p .Z

.

I1

prob:e -

-mo

Jr-I

45 kD

.I 29

kD

CREB--

-

[12 3 41L PKA

PF

t P

'2.PKA

Fig. 3. CREB binding to the TATCRE is dependent on phosphorylation. In (A) proteins from the oligodeoxynucleotide affinity purification of rat brain CREB were analysed by SDS-PAGE and silver staining. Aliquots of the starting material (lane 1), the flowthrough (lane 2) and the retained protein (lane 3) are shown. The marker proteins are shown at the left (lane M). Bandshift assays with PKA- or PP2A-treated brain CREB are shown in (B). Equivalent amounts of purified CREB were incubated with PKA alone, PP2A alone, or PP2A followed by PKA in the presence of okadaic acid to inhibit PP2A activity. All lanes in Figure 3B contain the same amount of purified CREB protein. Each of the three samples of treated CREB was mixed with 0.5, 2.5, 5 or 10 fmol (lanes 1-4, 5-8 and 9-12, respectively) of labelled TATCRE and incubated for 1 h on ice. The enzyme treatment

of CREB is shown at the bottom of the lanes.

kDa band (Figure 3A), for phosphorylation-induced binding to the TATCRE. We analysed binding activity of brain CREB which had been treated with either the catalytic subunit of PKA to fully phosphorylate the PKA site(s), or with protein phosphatase 2A (PP2A) to remove phosphates. CREB treated with PP2A showed little binding to the TATCRE as compared with the PKA-treated sample (Figure 3B). The bandshift activity after PP2A treatment could be recovered when the protein was rephosphorylated by PKA in the presence of okadaic acid, which inhibits PP2A. Therefore, as for the BC activity from rat liver, it appears that the phosphorylation state of CREB influences DNA binding to the TATCRE. A 43 kDa protein present in the CREB and BC fractions is the major substrate for the catalytic subunit of PKA in vitro The most probable explanation for the increased binding to the TATCRE with the liver BC and the brain CREB fractions after incubation with PKA is that phosphorylation of a protein present in both of those fractions mediates DNA binding. To identify the substrates for PKA, the purified BC and CREB fractions were phosphorylated in vitro using purified catalytic subunit of PKA and [-y-32P]ATP. The reaction products were resolved by SDS -PAGE and reveal that the only major substrate for PKA in each of these fractions was a 43 kDa protein (Figure 4A). This phosphoprotein comigrates with the immunoreactive protein shown in Western blot analysis (Figure 2A) and the purified protein in the silver stained SDS gel (Figure 3A), suggesting that they are identical. Additional phosphoproteins detected in the liver fraction were not consistently present in this assay and may be degradation products of the 43 kDa protein. When the A fraction was used as substrate for PKA, no phosphoproteins were seen, confirming that it is not a direct target of PKA (not shown).

Phosphorylation of Ser133 by PKA increases binding to the TATCRE Phosphoamino acid analysis of the major 43 kDa phosphoprotein confirmed that both BC and CREB were labelled only at serine residues upon phosphorylation with PKA in vitro (not shown). To determine the site(s) of phosphorylation, we performed tryptic digests and twodimensional peptide analyses of the 43 kDa phosphoproteins from the BC and CREB fractions. They each showed the same single tryptic phosphopeptide, following phosphorylation by the catalytic subunit of PKA in vitro, whose identity was confirmed by mixing experiments including a synthetic marker peptide predicted from the CREB sequence (Figure 4B). Thus, a single site (Serl33) on CREB from rat brain and BC from liver is phosphorylated by PKA in vitro. An otherwise identical marker peptide containing Ala in place of Ser at position 133 could not be phosphorylated by PKA, confirming the phosphorylation site (data not shown). To compare the residue(s) modified in vitro and in vivo, phosphorylation of the 43 kDa protein in response to forskolin was examined in FT02B hepatoma cells. Following phosphorylation in vivo, only a 43 kDa phosphoprotein was immunoprecipitated from soluble cellular proteins by the CREB antibodies and peptide analysis was performed. The same tryptic phosphopeptide was observed after phosphorylation in vivo as compared with 3339

M.Nichols et al.

i'ee TJP

Fig. 4. The major substrate of PKA in the purified BC and brain CREB fractions is a 43 kDa protein, and two-dimensional tryptic peptide mapping of the 43 kDa phosphoprotein identifies Serl33 as the PKA-dependent phosphorylation site. Aliquots of either brain CREB or liver BC fractions were phosphorylated by PKA using [-y-32P]ATP and reaction products were run in triplicate on a 10% SDS-polyacrylamide gel (A). The 43 kDa phosphoprotein from BC or CREB fractions was digested by trypsin following phosphorylation with PKA in vitro. Trypsin digests were subjected to phosphopeptide mapping as described in Materials and methods. Peptides were resolved in the horizontal dimension by electrophoresis and in the vertical dimension by ascending chromatography (B). An arrow marks the origin. A single phosphopeptide is observed with BC or CREB, phosphorylated in vitro by PKA and it migrates at an identical position as the phosphorylated marker peptide when mixed (CREB + Marker peptide). Phospholabelled CREB was immunoprecipitated from equal amounts of extracts of FTO2B cells which had been incubated for 4 h with then treated with forskolin (or solvent control) and subjected to phosphopeptide mapping (B). The single phosphopeptide observed migrated identically with the marker peptide in mixing experiments (not shown). CREB from liver is a multiply modified protein (C). An aliquot of the BC fraction from liver was labelled using PKA and [-y-32P]ATP, mixed with a 100-fold excess of untreated BC fraction and resolved by two-dimensional electrophoresis. The resulting gel was further analysed by Western blotting and the immunostained forms are shown in the panel labelled ab. The autoradiograph of the blotted proteins is shown in the panel labelled 32p. Immunostaining with anti-CREB antibodies reveals five forms migrating at 43 kDa. The 32p signal identifies those proteins which have been labelled in vitro by PKA. The CREB proteins shift by only one charge interval (as determined by isoelectric focusing gel markers) after PKA phosphorylation and are summarized schematically in the lower half of the figure. The cathode (+) and anode (-) for the isoelectric focusing gel are noted at the top of the figure.

purified CREB or marker peptide phosphorylated by PKA in vitro. The amount of phosphate at the Ser133 site increased after forskolin treatment of the cells (Figure 4B and Boshart et al., 1991). Therefore, the collective evidence from bandshift assays, antibody analysis, SDS gel migration, PKA phosphorylation and phosphopeptide mapping shows that the BC protein from liver is the same as CREB from brain. Multiple modified forms of CREB exist in liver To see whether multiple phosphorylated forms exist in vivo, we analysed CREB from liver by two-dimensional protein gel electrophoresis, followed by Western blot analyses. We found that there are five protein species of Mr 43 000 (labelled 1-5), differing by net charge increments of one and centred at pl = -5.8, which are recognized by the CREB antibodies (Figure 4C). After phosphorylation with PKA of purified CREB from liver, all five species (1-5) quantitatively moved one charge interval toward the cathode (+), consistent with increased negative charge from one added phosphate. This indicates that only one PKA site is present in CREB, in agreement with the tryptic peptide analysis (Figure 4B). Since all CREB forms (1-5) are substrates for the PKA phosphorylation, prior modification appears not to be required for phosphorylation at Serl33. Two forms of CREB (a and A) have been identified which differ only by an alternatively spliced a-exon, encoding 14 3340

amino acids (Yamamoto et al., 1990; Ruppert et al., 1992). The ca-specific amino acids contribute a net charge of +2, so CREBaz forms should migrate two charge units toward the anode (-), relative to the CREBA forms. This was confirmed using the a and A forms of CREB protein expressed in Escherichia coli (data not shown). With this in mind, forms 3-5 correspond to CREBA while forms 1-3 correspond to CREBa (Figure 4C). Hence, CREBA would comprise 80-90% of the CREB protein found in both liver and brain (proteins from both sources show identical 2-D patterns), consistent with mRNA ratios in liver (Ruppert et al., 1992) and brain (Yamamoto et al., 1990). Phosphorylation consensus sites for additional protein kinases are present N-terminal to the PKA site in CREB (Lee et al., 1990) and are not resolved in tryptic peptide analysis as the tryptic fragment containing these sites is larger than 100 amino acids. Surprisingly, the migration of the CREB forms in the 2-D gel analysis (Figure 4C) are not changed by extensive treatment with several phosphatases, arguing that these forms differ by something other than phosphates, or that the phosphates are extremely resistant to these phosphatases. However, phosphatases PP2A and PP1 are capable of completely reversing the migration change caused by PKA treatment in vitro (not shown). These modifications may include glycosylation as can be predicted from the protein sequence. In any case PKA phosphorylation at Serl33 does not require protein modification at other sites.

CREB binding to high and low affinity CREs

In vivo footprinting reveals enhanced binding at the TATCRE but not at the PEPCK CRE after forskolin treatment We have seen previously in hepatoma cells that binding to the TATCRE was increased by PKA stimulation (Weih et al., 1990). We wished to extend our analysis to another CRE that is active in hepatoma cells and therefore analysed protein binding in vivo at the CRE of the PEPCK gene, with and without stimulation of PKA by forskolin. The CRE of the PEPCK gene is well characterized (Quinn et al., 1988) and is a representative of the symmetric CRE type (see below). As well, PEPCK gene transcription is induced by increased intracellular cAMP levels in hepatoma cells (Wynshaw-Boris et al., 1984; Stewart and Schutz, 1987; Sasaki and Granner, 1988). As shown in Figure 5, the TAT and PEPCK CREs clearly show a footprint in FT02B hepatoma cells, relative to the DMS protection pattern in the fibrosarcoma cell line XC, which does not express TAT or PEPCK and does not show a footprint even after forskolin treatment (not shown). The DMS protection patterns at the PEPCK CRE in uninduced (un) and forskolin induced (cAMP) hepatoma cells appear to be identical. This is in contrast to the increased occupancy at the TATCRE after forskolin induction (Figure 5A; Weih et al., 1990). We also examined protein binding at the PEPCK CRE in FT02B hepatoma cell derivatives (clones # 17 and # 21) which have a strongly repressed level of PKA activity, resulting from overexpression of the PKA regulatory subunit RIct (Boshart et al., 1991). In these cells, the footprint at the PEPCK CRE and the TATCRE nearly disappears, confirming that binding to both CREs is sensitive to PKA activity, and correlating with the strongly reduced expression of the PEPCK and TAT genes in these cell lines (Boshart et al., 1991).

High and low affinity CREs show increased binding by CREB after PKA phosphorylation The observation that binding of CREB to the TATCRE is enhanced after PKA activation in vivo or in vitro prompted us to examine other CREs in bandshift assays, particularly as the PEPCK CRE did not show increased binding in vivo when hepatoma cells were stimulated with forskolin (Figure 5). Since the most striking difference between the TATCRE and consensus CREs such as the somatostatin CRE is that the former is asymmetric (ctgCGTCA) and the latter symmetric (TGACGTCA), we analysed additional well characterized CREs from these two classes (Figure 6). Two CREs upstream of the urokinase-type plasminogen activator gene (uPA-A and uPA-B; von der Ahe et al., 1990) and a CRE from the proenkephalin gene (ENK; Comb et al., 1988) are in the asymmetric CRE category, while the somatostatin (SOM; Montminy and Bilezikjian, 1987), ctchorionic gonadotropin (oaCG; Delegeane et al., 1987) and the fibronectin (FIB; Dean et al., 1989) genes have symmetric CREs. We also designed a mutant with three nucleotide changes in the symmetric SOMCRE which changed it to the asymmetric class (SOM 5/8). This collection of nine sequences was then tested in bandshift assays with CREBa protein from an E. coli expression clone, thereby assuring homodimer CREB protein (Figure 6). Similar results were found using protein encoded by the CREBA cDNA in either E. coli or baculovirus expression systems (data not shown). Clearly, a difference in binding

PEPCK CRE

C.T-- v~'..

X -~ C

'-

_3

Sel-es

-- P" !l